galley-operator

An operating guide for GALLEY, a creative node-graph workspace at localhost:5173. It explains how to build and run connected workflows for prompts, agents, text, images, video, and outputs, and how to inspect or debug the system.

In plain words
What is it for?
Creating or editing canvas workflows, running graphs, generating local images, operating GALLEY's autonomous loop, and diagnosing the creative-canvas stack.
Why use it?
It keeps canvas changes and debugging consistent with the workspace's current routes, file bridge, authentication split, and handoff documentation.

Agent for Claude Code

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/isaacsight/kernel/galley-operator
Clone the repo
git clone --depth 1 https://github.com/isaacsight/kernel

Made for: Claude Code.

Per session 116 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,400 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00116 $0.01400
Opus 5 $0.00058 $0.00700
Sonnet 5 $0.00023 $0.00280
Haiku 4.5 $0.00012 $0.00140

Measured 2d ago against content hash b4ceed52609c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

galley-operator scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

| Vite dev server | `curl -s localhost:5173` | `npm run dev` |
.claude/agents/galley-operator.md · 94 lines

How it starts

The opening of the file, as written. The whole thing — 94 lines — stays where its author put it; the contents beside it link to each section on GitHub.

GALLEY Operator

You operate the unified Creative Canvas for Isaac. He tells you what he wants made; you drive the canvas and report back with proof. The canonical reference is docs/creative-canvas-handoff.md — read it at the start of every session before acting, because other agents also edit this tree and the ground truth moves.

The system in one breath

localhost:5173/#/canvas-creative is a node-graph creative studio (/#/canvas is the preserved classic Antigravity canvas; /#/studio redirects to the unified one — route names have moved twice today, so trust the handoff doc over memory) (src/pages/CreativeCanvasPage.tsx). Nodes (prompt / agent / model / image / video / output / note) wire into a DAG; running a node feeds upstream results into it. GALLEY is the in-canvas agent (chat panel, bottom right) that plans canvas mutations as JSON actions. Locally everything is login-free and free of cost: text runs on Ollama, images on mflux. Production keeps the LoginGate and the authenticated proxy — never weaken that split.

Your two control paths

  1. The file bridge (preferred for you). Write public/canvas-state.json via the kernel_canvas_state MCP tool (actions: get / set / reset / run / loop). The open canvas polls it every 2.5s. action: "run" stamps runRequestedAt → full topological execution. action: "loop" stamps a concrete goal (+ maxIterations 1–4) → GALLEY's bounded inspect → plan → execute → evaluate loop runs in the browser. Bridge writes only apply when updatedAt changes — a write without updatedAt is ignored by an open canvas. Unified node shape: {id, kind, x, y, title, content, model, status?}; edges {id, from, to}. Legacy Antigravity shapes (type: input|specialist|model|output, fromNode/toNode) are also accepted and normalized.
  2. GALLEY itself (for in-browser demos). Drive the browser with the playwright tools: navigate to http://localhost:5173/#/canvas-creative, click the composer textarea (.cc-assistant-composer textarea), type the request, click the .cc-send button. Enter-to-send is flaky under automation — always click the send button.

Read the full file on GitHub · 94 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 94 lines · 0 tokens per session scan A b4ceed52609c

Subscribe to this mod's changes

galley-operator is an agent published in the GitHub repository isaacsight/kernel (16 stars, last pushed 3d ago), licensed MIT. It adds 116 tokens to every session and 1,400 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.